| Message ID | CAPgLHd-+DNxxVHsXiJpk2KFk8mzrQUkwaYPUFeWHyAmz-H6=4Q@mail.gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Mar 26, 2013 at 02:42:47PM +0800, Wei Yongjun wrote: > From: Wei Yongjun <yongjun_wei@trendmicro.com.cn> > > sizeof() when applied to a pointer typed expression gives the > size of the pointer, not that of the pointed data. > This fix isn't right. "buf" is a char pointer. I don't know what this code is doing. Instead of sizeof(*buf) it should be something like "buflen", "msg[i].len", "msg[i].len + 1" or "msg[i].len + 3". I'm not sure which is correct here or what it's doing, sorry. regards, dan carpenter -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Mar 26, 2013 at 10:04:15AM +0300, Dan Carpenter wrote: > On Tue, Mar 26, 2013 at 02:42:47PM +0800, Wei Yongjun wrote: > > From: Wei Yongjun <yongjun_wei@trendmicro.com.cn> > > > > sizeof() when applied to a pointer typed expression gives the > > size of the pointer, not that of the pointed data. > > > > This fix isn't right. "buf" is a char pointer. I don't know what > this code is doing. Instead of sizeof(*buf) it should be something > like "buflen", "msg[i].len", "msg[i].len + 1" or "msg[i].len + 3". It should be "msg[i].len + 1", I think. On the line before it writes buflen bytes to the hardware. Then it clears the transfer buffer and reads "msg[i].len + 1" bytes from the hardware. Then it saves the memory, except for the first byte, in msg[i].buf. So it should clear "msg[i].len + 1" bytes so that the old data isn't confused with the data that we read from the hardware. regards, dan carpenter -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue March 26 2013 08:35:57 Dan Carpenter wrote: > On Tue, Mar 26, 2013 at 10:04:15AM +0300, Dan Carpenter wrote: > > On Tue, Mar 26, 2013 at 02:42:47PM +0800, Wei Yongjun wrote: > > > From: Wei Yongjun <yongjun_wei@trendmicro.com.cn> > > > > > > sizeof() when applied to a pointer typed expression gives the > > > size of the pointer, not that of the pointed data. > > > > > > > This fix isn't right. "buf" is a char pointer. I don't know what > > this code is doing. Instead of sizeof(*buf) it should be something > > like "buflen", "msg[i].len", "msg[i].len + 1" or "msg[i].len + 3". > > It should be "msg[i].len + 1", I think. Yes, that's correct. 'buf' used to be a local array, so the memset was fine. I changed it to an array that was kmalloc()ed but forgot about the memset. I never noticed the bug because the sizeof the message is typically quite small, certainly smaller than sizeof(pointer) on a 64-bit system. Wei Yongjun, can you post a new patch fixing this? Thanks, Hans > > On the line before it writes buflen bytes to the hardware. Then > it clears the transfer buffer and reads "msg[i].len + 1" bytes from > the hardware. Then it saves the memory, except for the first byte, > in msg[i].buf. > > So it should clear "msg[i].len + 1" bytes so that the old data isn't > confused with the data that we read from the hardware. > > regards, > dan carpenter > > -- > To unsubscribe from this list: send the line "unsubscribe linux-media" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Hans and Dan Carpenter, On 03/26/2013 04:18 PM, Hans Verkuil wrote: > On Tue March 26 2013 08:35:57 Dan Carpenter wrote: >> On Tue, Mar 26, 2013 at 10:04:15AM +0300, Dan Carpenter wrote: >>> On Tue, Mar 26, 2013 at 02:42:47PM +0800, Wei Yongjun wrote: >>>> From: Wei Yongjun <yongjun_wei@trendmicro.com.cn> >>>> >>>> sizeof() when applied to a pointer typed expression gives the >>>> size of the pointer, not that of the pointed data. >>>> >>> This fix isn't right. "buf" is a char pointer. I don't know what >>> this code is doing. Instead of sizeof(*buf) it should be something >>> like "buflen", "msg[i].len", "msg[i].len + 1" or "msg[i].len + 3". >> It should be "msg[i].len + 1", I think. > Yes, that's correct. > > 'buf' used to be a local array, so the memset was fine. I changed it to an > array that was kmalloc()ed but forgot about the memset. I never noticed > the bug because the sizeof the message is typically quite small, certainly > smaller than sizeof(pointer) on a 64-bit system. > > Wei Yongjun, can you post a new patch fixing this? Thanks very much, I will send the v2 of this patch soon. Regards, Yongjun -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/staging/media/go7007/go7007-usb.c b/drivers/staging/media/go7007/go7007-usb.c index 0823506..7219ae0 100644 --- a/drivers/staging/media/go7007/go7007-usb.c +++ b/drivers/staging/media/go7007/go7007-usb.c @@ -1035,7 +1035,7 @@ static int go7007_usb_i2c_master_xfer(struct i2c_adapter *adapter, buf, buf_len, 0) < 0) goto i2c_done; if (msgs[i].flags & I2C_M_RD) { - memset(buf, 0, sizeof(buf)); + memset(buf, 0, sizeof(*buf)); if (go7007_usb_vendor_request(go, 0x25, 0, 0, buf, msgs[i].len + 1, 1) < 0) goto i2c_done;